What if your AI could connect the dots without you telling it what is related?
Most knowledge tools put the connection work on you. Tag this. Link that. File it in the right folder. Organize it or lose it.
But the whole point of using AI is to spend less time on busywork, not more.
Second Brain builds its knowledge graph automatically. When you save a memory, the system extracts the people, projects, decisions, and concepts mentioned. It scans the existing graph for related entries and creates links where it finds meaningful overlap.
A few examples of what this surfaces in practice:
You store a note about a product decision. The system links it to a previous discussion about the same feature, a person mentioned in both, and an older decision that this one partly reverses.
You store a meeting summary. It links to the project it references, the last meeting about the same topic, and an open question that this meeting answered.
The links are not perfect. Some are too broad. Some miss things. You can remove or correct them. But the default behavior is that the graph grows on its own as you use it, and recall gets better the more you add.
We launched the desktop app today. Five minutes to set up, no terminal required: https://www.thesecondbrain.dev
We are live on Product Hunt: https://www.producthunt.com/posts/second-brain-for-mac-and-windows
After a day of using an AI memory tool, what types of automatic connections would be most valuable to you?
Replies
the risk I keep coming back to with automatic linking is false-positive fatigue - if you have to review and prune a chunk of "meaningful overlap" links every time you add a note, that's just busywork wearing a different hat, tagging by another name. what's the actual precision looking like once your graph gets into the thousands-of-entries range, where two unrelated things sharing a person's name or a generic project word starts producing noisy links?
the connection I'd actually want most is contradiction detection, not just relation - not "here's a related note" but "this note conflicts with something you said three months ago, which one's current." that's a different job than similarity linking though. separate question from the noise one Omri raised: what about connections that cross a boundary you didn't intend, like a work decision auto-linking to a personal note because the same person's name shows up in both? similarity doesn't know those live in different mental buckets for you. is there any notion of context/namespace in the graph, or is everything one flat pool once it's stored?